Surgical decision-making in the management of malignant bowel obstruction (MBO) is challenging due to a complex interplay of clinical considerations, goals of care, and a paucity of evidence in the literature to guide practice.1,2,3 As a common sequela of advanced cancer, particularly gastrointestinal or ovarian cancer, MBO often causes abdominal pain, nausea, vomiting, bloating, and appetite loss.2,4 Nonoperative treatment with supportive care, endoscopic stenting, or gastrostomy tube placement is often the first line of treatment.3 However, with persistent or complete obstructions, surgical intervention with bowel resection, bypass, and/or ostomy creation is typically considered.3,5 Yet many patients with MBO have disseminated cancer and are poor surgical candidates due to large disease burden and malnutrition.2 Therefore, these patients often have a narrow therapeutic margin for surgical palliation of MBO.5 Existing evidence demonstrates that operative management often alleviates obstructive symptoms and allows resumption of a diet, but carries high risk of mortality, serious complications, recurrence, and lengthened hospitalization.2,4,6

The decision to operate in the setting of MBO is nuanced, and developing an understanding of individual patient goals is essential.1 Focused discussions addressing prognosis, end-of-life goals, and risks of surgery are central to shared decision-making.1,7 Generally, these operations are not performed for curative intent, with symptom relief and quality of life prioritized over prolongation of life.1,7 Although palliative care and hospice services are important adjuncts to the achievement of these goals, evidence points to lower utilization and delays in initiation of these services for Black, Hispanic, and Asian patients.8,9,10,11,12,13,14,15

Additional disparities have been identified for Black patients undergoing cancer care, including an underuse of surgery and chemotherapy,16,17,18 and a higher cancer mortality rate than for other ethnoracial groups.19,20 Potential contributors to these disparities may include Black patients presenting at an advanced disease stage, having a higher comorbidity burden, being seen less often by high-volume surgeons, having systemically limited access to care, and experiencing higher social vulnerability.21,22,23,24,25

Growing evidence shows disparities among ethnoracial groups encompassing oncologic management and palliative care utilization. When surgical palliation is considered for patients with MBO, understanding the relationship between race/ethnicity, management options, and expected outcomes is vital. Although race and ethnicity are imprecise social constructs, in this context, they can aid in identifying modifiable systemic disparities.26,27

The primary aim of this study was to compare ethnoracial differences in major morbidity after surgery, defined as complications that require invasive intervention and risk organ failure or death.28 The secondary aims included comparing ethnoracial differences with respect to preoperative factors, operative characteristics, overall postoperative morbidity, postoperative mortality, need for reoperation, hospital length of stay (LOS), discharge disposition, and readmission. We hypothesized that Hispanic, Asian, and Black patients have significantly worse outcomes after surgery for MBO than White patients.

Methods

Study Design and Patient Population

This retrospective study included data collected from 2010 through 2019 from the registry of the American College of Surgeons National Surgical Quality Improvement Program (NSQIP), a multi-institutional quality improvement registry that collects perioperative data through 30 days after major surgical procedures at more than 700 participating sites.29 Patients were identified who had disseminated cancer and a primary postoperative diagnosis of intestinal obstruction from a cause other than adhesions (ICD-9 codes 560.89, 560.9; ICD-10 codes K56.60, K56.69), similar to previously published methods.30

The consensus definition of MBO for clinical trials requires clear evidence of bowel obstruction, obstruction beyond the ligament of Treitz, and either incurable intra-abdominal primary cancer or non-intra-abdominal primary cancer with intraperitoneal disease.31 In this definition, the location of MBO is specified because obstructions proximal to the ligament of Treitz are primarily treated endoscopically.31 However, the NSQIP database includes only patients undergoing surgical procedures and does not specify MBO location, which cannot always be determined based on the procedure performed.

The patients were categorized by race/ethnicity. The study first grouped all the patients with Hispanic ethnicity regardless of race, then grouped the non-Hispanic patients as non-Hispanic White, non-Hispanic Black or African American, or non-Hispanic Asian, referenced as White, Black, and Asian, respectively. Native American or Alaska Native and Native Hawaiian or Pacific Islander populations were excluded from the analysis due to a small sample size. Patients with unknown race and ethnicity were also excluded.

The Duke University Health System Institutional Review Board determined the study to be exempt from review (Pro00107261). This study was conducted in accordance with the Checklist to Evaluate the Science of Surgical Database Research, the NSQIP-specific guide, and the RECORD statement.32,33,34

Variable Definitions

Independent patient variables were grouped for clinical relevance, as subsequently described, to facilitate interpretation and analysis. The variables included were age (those recorded as “90+” were changed to 90 for analysis), sex, direct admittance versus transferred (from other health care facilities), body mass index (BMI: underweight [<18.5 kg/m2], normal [18.5–24.99 kg/m2] overweight [25–29.99 kg/m2], or obese [≥30 kg/m2]), American Society of Anesthesiologists (ASA) classification (1–2 [no disturbance–mild disturbance] or 3–5 [severe disturbance–moribund]), functional status (independent or dependent [partially or totally]), weight loss (>10% during 6 months), immunosuppression (long-term with corticosteroids or other immunosuppressive agents), diabetes (with or without insulin), hypertension (requiring pharmacologic treatment), congestive heart failure (CHF), smoking (cigarettes within 12 months previously), dyspnea (with moderate exertion or at rest), severe chronic obstructive pulmonary disease (COPD), ventilator dependence (within 48 h after surgery), ascites, acute renal failure (within 24 h after surgery), dialysis (within 2 weeks after surgery), open wound, bleeding disorder, preoperative red blood cell (RBC) transfusion (within 72 h after surgery), and preoperative sepsis (systemic inflammatory response syndrome, sepsis, or septic shock). Preoperative lab values from the most recent lab drawn within 90 days before the procedure, including albumin (g/dL), creatinine (mg/dL), bilirubin (mg/dL), hematocrit (%), and white blood cell (WBC) count (×109 cells/L), were used.

Preoperative LOS indicated the days from hospital admission to operation. All current procedural terminology (CPT) codes, including “primary,” “concurrent,” and “other” CPT categories, were evaluated to group procedure types. Procedures involving the small bowel, large bowel, or stomach were identified, and bowel procedures were sub-categorized as resection/bypass, ostomy, or other.

Outcome Definitions

The primary outcome was 30-day major morbidity, defined as Clavien-Dindo classification grade 3 or 4 disease requiring invasive intervention with risk of organ failure or death.28 Major morbidity included deep or organ space surgical-site infection (SSI), wound dehiscence, pulmonary embolism (PE), ventilation longer than 48 h after surgery, unplanned reintubation, acute renal failure, progressive renal insufficiency, sepsis, septic shock, myocardial infarction (MI), cardiac arrest requiring cardiopulmonary resuscitation (CPR), and cerebrovascular accident (CVA).

The secondary outcomes were 30-day mortality, overall morbidity, reoperation (unplanned return to the operating room), unplanned readmission, total LOS (days from admission to discharge), and discharge destination. Overall morbidity included all major complications in addition to superficial SSI, pneumonia, urinary tract infection (UTI), intra- or postoperative blood transfusion, and deep venous thrombosis (DVT) requiring therapy. Discharge destination was dichotomized as discharge to home indicating “home” or “facility which was home,” or as not discharged to home.

Statistical Analysis

Descriptive statistics were performed to compare ethnoracial differences in baseline characteristics. For all analyses, White race was the reference group. Categorical variables were compared using chi-square or Fisher’s exact test as appropriate and reported as frequencies or proportions. Continuous variables were compared using Wilcoxon rank sums, with median and interquartile ranges (IQR) reported. Multivariable logistic regression models were created to assess ethnoracial differences in major morbidity and overall morbidity while controlling for potential confounders.

Variables were selected for inclusion in the model by first assessing the univariate significance of all preoperative and operative characteristics for the outcome of major morbidity to identify those with a significance level of p lower than 0.05 for consideration of inclusion in the model. Additionally, forward stepwise selection using all preoperative and operative characteristics was performed to ensure that no important variables were overlooked. The final selection of variables for the model was based on these results in addition to clinical knowledge of which variables are known to have an impact on morbidity after surgery to ensure that the included variables were clinically relevant.

Results were reported using odds ratios (ORs) and 95% confidence intervals (95% CIs). All statistical tests were two-sided, with a p value lower than 0.05 considered statistically significant. Analyses were performed using JMP version 15.1, SAS Institute, Cary, North Carolina.35

Results

Patient Characteristics

From 2010 to 2019, 173,960 patients had disseminated cancer, and 3211 (1.8%) of these patients had a postoperative diagnosis of bowel obstruction. Eight patients (0.2%) were removed from the analysis because the procedures performed would not be indicated for MBO (e.g., hip arthroplasty or exploration of the retroperitoneum). Of the remaining 3203 patients, 441 (13.8%) were excluded because their ethnoracial group was unknown or had a sample size that was too small for analysis.

The final cohort for the analysis was comprised of 2762 patients: 2081 White patients (75.3%), 407 Black patients (14.7%), 159 Hispanic patients (5.8%), and 115 Asian patients (4.2%). Figure 1 outlines the steps for patient selection.

Fig. 1
figure 1

Steps of selection for study inclusion

Full size image

The demographic and preoperative characteristics by ethnoracial group are described in Table 1. The overall cohort had a median age of 65 years (IQR, 56–73 years) and was 54.7% female (n = 1510). Compared with other ethnoracial groups, White patients were relatively older (66 years; IQR, 57–74 years) and more likely to have been transferred from another institution (n = 403, 19.4%). Black patients were more likely than White patients to be underweight (10.2% vs 7.0%; p = 0.0295), to be obese (25.0% vs 19.0%; p = 0.0060) or to have diabetes (17.9% vs 12.7%; p = 0.0051), hypertension (54.1% vs 44.7%; p = 0.0006), ventilator dependence (1.7% vs 0.5%; p = 0.0183), acute renal failure (3.2% vs 1.1%; p = 0.0012), or preoperative RBC transfusion (8.1% vs 4.1%; p = 0.0005). Black patients also were more likely to have a higher median level of creatinine (0.88 mg/dL [IQR, 0.7–1.18 mg/dL] vs 0.8 mg/dL [IQR, 0.6–1.03 mg/dL]; p < 0.0001) or bilirubin (0.6 mg/dL [IQR, 0.4–0.9 mg/dL] vs 0.5 mg/dL [IQR, 0.4–0.8 mg/dL]; p = 0.0359), and a lower median level of hematocrit (31.6% [IQR, 27.95–35.5%] vs 33.3% [IQR, 29.8–37.2%]; p < 0.0001). Hispanic and Asian patients were similar to White patients in terms of preoperative characteristics and labs, often having fewer comorbidities. There were no ethnoracial differences in sex, ASA classification, dependent functional status, weight loss, immunosuppression, CHF, ascites, dialysis, open wounds, sepsis, preoperative albumin, or postoperative WBC count.

Table 1 Demographic and preoperative characteristics by ethnoracial groupa
Full size table

Operative characteristics by ethnoracial group are described in Table 2. Overall, 42.4% of the patients had a small bowel procedure, 50.4% had a colorectal procedure, and 7.2% had a gastric procedure, with 53.2% receiving an ostomy and 4.8% receiving a gastrostomy. Emergency procedures were more likely for Black patients (37.1% vs 30.0%; p = 0.0046) and Asian patients (39.1% vs 30.0%; p = 0.0381) than for White patients. Asian patients also were more likely to have a small bowel procedure (52.2% vs 42.1%; p = 0.0334) and less likely to have a colorectal procedure (39.1% vs 51.2%; p = 0.0116) than White patients. The analysis found no ethnoracial differences in preoperative LOS (median, 2 days; IQR, 1–5 days), wound classification (clean/contaminated most common, 67.6%), or operation time (median, 102 min; IQR, 69–158 min).

Table 2 Operative characteristics by ethnoracial groupa
Full size table

Postoperative Outcomes

The results from the univariate analyses of postoperative outcomes are described in Table 3. The overall 30-day mortality rate was 12.6%. Additionally, 23.0% of the patients had at least one major complication, with 41.5% of the patients having at least one overall complication and 23.1% having two or more overall complications. The most common complications were RBC transfusion (16.0%), sepsis (8.3%), superficial SSI (7.2%), organ space SSI (7.1%), pneumonia (5.5%), and septic shock (4.7%). The median LOS was 11 days (IQR, 7–17 days), with a median time from operation to discharge of 7 days (IQR, 5–12 days). The unplanned reoperation rate was 7.8%, and the unplanned readmission rate was 17.6%. Most of the patients (72.8%) were discharged to home.

Table 3 Univariate analysis of 30-day postoperative outcomes by ethnoracial groupa
Full size table

A comparison of outcomes by ethnoracial group showed that Black patients were more likely than White patients to have at least one major complication (28.5% vs 21.8%; p = 0.0031), at least one overall complication (47.4% vs 40.4%; p = 0.0087), and two or more overall complications (23.1% vs 17.2%; p = 0.0044). The complications significantly more likely among Black patients were prolonged ventilation (6.6% vs 3.8%; p = 0.0112), progressive renal insufficiency (3.2% vs 1.1%; p = 0.0012), cardiac arrest (3.0% vs 0.7%; p = 0.0005), and RBC transfusion (19.4% vs 15.2%; p = 0.0329).

Black patients also had a longer median LOS (12 days [IQR, 8–19 days] vs 10 days [IQR, 7–17 days]; p = 0.0007) and a longer time from operation to discharge (9 days [IQR, 5–13 days] vs 7 days [IQR, 5–11 days]; p = 0.0006). Black patients were less likely to be discharged to home (67.6% vs 73.0%; p = 0.0315) and more likely to have an unplanned readmission (17.1% vs 12.9%; p = 0.0266).

Hispanic patients had a lower mortality rate than White patients (6.3% vs 13.1%; p = 0.0130), and a longer LOS (12 days [IQR, 8–18 days] vs 10 days [IQR, 7–17 days]; p = 0.0313). Asian patients did not differ significantly in postoperative outcomes from White patients.

The results of the multivariable logistic regression analyses are displayed in Table 4. After adjustment for clinically relevant preoperative and operative characteristics, the multi-categorical ethnoracial variable was not significantly associated with major morbidity (p = 0.1061) or overall morbidity (p = 0.1487). However, Black patients had significantly higher odds of experiencing a major complication (OR, 1.42; 95% CI, 1.07–1.88; p = 0.0152) or overall complication (OR, 1.34; 95% CI, 1.05–1.73; p = 0.0208) than White patients.

Table 4 Multivariable logistic regression model for odds of postoperative major and overall morbidity by ethnoracial groupa
Full size table

Discussion

This study demonstrated that Black patients had a significantly higher risk of major and overall 30-day postoperative morbidity, a longer LOS, a lower probability of discharge to home, and higher rates of unplanned readmission after surgery for MBO than White patients. The complications more likely for Black patients were prolonged ventilation, progressive renal insufficiency, cardiac arrest, and RBC transfusion. The 30-day mortality rates did not differ significantly between Black and White patients.

Many factors contribute to ethnoracial disparities in surgical outcomes, with differences in preoperative health as one potential cause.22 Black patients in this study had greater preoperative comorbidity and more emergency procedures than White patients, suggesting suboptimal medical management of chronic illness before surgery, likely secondary to impaired health care access and inequities in care received, creating a more emergent need for intervention and increased surgical risk at the time of presentation.36,37,38 Importantly, differences in morbidity for Black versus White patients persisted after controlling for comorbidities and emergency procedures in the multivariable model, suggesting that additional systemic inequalities are contributing.

Disparities in operative management and outcomes for Black patients have been identified in many other studies, including longer preoperative LOS,39 fewer minimally invasive procedures,40 more debilitating surgery,41 longer total LOS,42,43,44 higher rates of major complications,36 and higher mortality rates.42,43 These findings highlight the necessity of addressing potential disparities in patient care. They are particularly salient in the context of palliative interventions, such as surgery for MBO performed to alleviate symptoms and improve quality of life.1,7

With a high likelihood of major complications, clear communication of postoperative risks and understanding of patient goals are crucial.2,45 Patients often prioritize maximizing time at home toward the end of life, making outcomes such as prolonged LOS and lower likelihood of discharge to home for Black patients an important consideration.45,46 In addition to addressing the higher rates of preoperative comorbid conditions observed for Black patients, identification of other opportunities to improve surgical outcomes is needed.

Another contributing factor may be that Black patients are more likely to undergo aggressive interventions at the end of life8,9,10,11,12,13,14,15 and less likely to have a do-not-resuscitate (DNR) order, even when they have a more severe illness.47,48 The reasons for this may include decreased palliative care utilization; impaired communication due to differences in language, culture, religion, or spirituality; overly optimistic beliefs about life-sustaining treatment; and/or provider mistrust.49,50,51

Implicit racial bias also can influence the way providers communicate with patients and may ultimately affect patient satisfaction, trust, and health-related behaviors.52 For example, physicians may hold the belief that Black patients with end-stage cancer are more likely than White patients to want potentially life-extending therapies,53 and may be less likely to discuss prognosis during palliative care conversations with Black and Latino patients who have advanced cancer.8 Additionally, Black patients have been shown to have a 30% higher likelihood of operative delay for small bowel obstruction,39 and are more likely to experience delays in initiation of cancer treatment for many common solid tumors.54

Differential treatment by physicians based on race, even if unintentional, may explain why Black patients were more likely to receive surgery for MBO with more preoperative comorbidities and a greater severity of illness. Therefore, it is important for providers to reflect consistently on how their biases could potentially have an impact on communication about goals of care, clinical management recommendations, timeliness of treatment, and other facets of cancer care.

Quality of care also varies widely by specific surgeon or hospital. Many factors shown to be associated with improved outcomes, such as having surgery performed by a specialist,55 receiving minimally invasive surgery,56 engaging in palliative care,57 and having surgery at a high-volume hospital,25 are less likely for Black patients or at hospitals serving larger proportions of ethnoracial minorities. Decreasing racial disparities in surgical mortality has been shown over time, but not for hospitals serving higher proportions of Black patients, emphasizing the need for more focus on these institutions.58

Surgical outcomes are influenced not only by care provided at the hospital, but also by a complex interplay of social determinants of health, including the resources to which patients have access.59 The impact of social vulnerability (e.g., poor access to educational opportunities, socioeconomic disadvantage, and language barriers) on postoperative outcomes is most pronounced for Black patients.60 Those living in communities with higher social vulnerability have higher morbidity, greater mortality, and longer LOS after resection for cancer,61 as well as disproportionately decreased hospice use.9 Socioeconomic factors and insurance status have been shown to confound survival differences between Black patients and White patients with colorectal cancer.62,63 Controlling for measures of social vulnerability was not feasible in the context of the current study, so such factors could explain the observed differences in outcomes by ethnoracial group.

In this study, Hispanic patients had a longer LOS, but all other outcomes were similar or superior to those of White patients. This difference may be explained in part by lower rates of preoperative comorbidities for this cohort. Findings have shown Hispanic patients to have lower morbidity and mortality due to cancer than White patients, but “Hispanic” encompases a heterogeneous group, and outcomes may differ significantly based on country of origin, immigration status, and language fluency.64 Outcomes did not differ significantly between White and Asian patients. This may have been due to the smaller sample of Asian patients and consequently a lower statistical power for detecting the rarer complications. Asian Americans have been understudied with respect to disparities research and comprise a heterogeneous group with potential differences in outcomes depending on which sub-populations are analyzed.65,66 Further research is needed for both Hispanic and Asian populations to identify disparities and associated drivers corresponding to specific sub-populations.

This study had several limitations. First, using ethnoracial categories creates mutually exclusive groups for a social construct that in reality is much more fluid and imprecise.27 However, this is a common approach to identification of systemic disparities and opportunities for intervention.

Second, the cohorts of Asian and Hispanic patients were the smallest in this study, and the database did not allow for analysis by sub-population such as ancestral country of origin. Thus, further investigation of these groups and other ethnoracial groups is needed.

Third, several other patient characteristics were not included in the NSQIP database that could contribute to poor outcomes or confound the results (e.g., hospital-level or surgeon-level clustering, education level, geographic location, socioeconomic factors, primary cancer type, tumor histology, driver mutations, or lesbian, gay, bisexual, transgender, queer, and questioning (LGBTQ+) identity. Future studies are needed to further analyze the impact of these factors.

Fourth, as described in the “Study Design and Patient Population” section, due to the retrospective nature of the NSQIP database, the ideal inclusion criteria for the diagnosis of MBO could not be applied.31

Fifth, the NSQIP database does not allow insight into more nuanced factors influencing decision-making such as goals-of-care discussions, patient satisfaction, or symptoms experienced. This limitation is common for large database studies of MBO.4,6,30,67 Finally, the NSQIP database is limited to a 30-day window, and hence outcomes outside of that time frame are not available for study.

In conclusion, Black patients had a higher risk of major and overall morbidity after surgery for MBO than White patients, independent of clinically relevant preoperative and operative characteristics. Black patients had mortality rates similar to those of White patients, but had higher rates of preoperative comorbidities, emergency procedures, prolonged LOS, and unplanned readmission, and a lower probability of being discharged to home. These findings have an important impact on surgical decision-making for MBO because any complications or prolonged hospitalization toward the end of life are major considerations. Differences in outcomes for Black patients have potential contributing factors including barriers to palliative care, implicit racial biases affecting communication, higher rates of comorbidities, greater social vulnerability, and clustering of lower-quality care at hospitals serving primarily Black patients. These disparities were not observed for Hispanic or Asian patients, although further research is needed to ensure that all opportunities to improve care are identified and addressed for these populations, as well as for other underrepresented groups. Future analyses controlling for potentially confounding variables such as socioeconomic status, education level, and tumor histology are indicated to identify underlying causes for these inequities. Surgeons should be mindful of the factors that have an impact on outcomes by ethnoracial group and how these can be mitigated to provide equally high-quality care across groups.